Improving similarity-driven library design: customized matching and regioselective feature trees.

Area of Science:

• Computational chemistry
• cheminformatics
• Drug discovery

Background:

• Reduced graph descriptors like feature trees are vital for analyzing molecular arrangements in fragment space search and library design.
• The LoFT (Library of Focused Trees) approach utilizes feature trees for novel focused library design.
• Existing feature tree methods struggle with regioisomeric distinctions and lack user-guided matching, leading to information loss and suboptimal alignments.

Purpose of the Study:

• To address the limitations of feature tree descriptors in distinguishing regioisomeric substructures.
• To introduce user-defined matching capabilities to improve the relevance of molecular alignments.
• To enhance the LoFT approach for more effective focused library design.

Main Methods:

• Modified feature tree generation to recognize diverse arene substitution patterns.
• Developed a customized matching procedure enabling users to specify allowed matching regions.
• Re-evaluated library design scenarios previously used for LoFT to assess improvements.

Main Results:

• The modified feature tree generation successfully differentiates various arene substitution patterns.
• Customized matching allows for more precise and user-relevant alignments.
• Improvements in library design were observed through the re-evaluation of design scenarios.

Conclusions:

• The enhanced feature tree descriptor overcomes limitations in regioisomer differentiation and alignment specificity.
• These improvements lead to more accurate and efficient focused library design.
• The refined approach offers greater control and precision in computational drug discovery efforts.